|Home | About | Journals | Submit | Contact Us | Français|
In the United States, a woman's health insurance coverage is largely determined by her employment and marital roles. This research evaluates competing hypotheses regarding how the combination of employment and marital roles shapes insurance coverage among Mexican-origin, non-Hispanic white, and African-American women. We use data from the 2004 and 2006 March Supplements to the Current Population Surveys. Results show these roles largely substitute for each other among non-Hispanic white and African-American women, although marriage generally increases the odds of coverage slightly more than employment among non-Hispanic white women. In contrast, these roles cumulatively increase the odds among Mexican-origin women. Yet, neither employment, nor marriage, nor their combination assures their coverage. Married Mexican-origin women are particularly disadvantaged. As women increasingly spend a smaller fraction of their lives married, and as benefit-rich occupations are the purview of few women, stable and equitable coverage may require a universal health insurance system.
In the United States, a woman's health insurance coverage is largely determined by her employment and marital roles. During the second half of the twentieth century, women increasingly entered the workforce and many now assume full responsibility for their health insurance coverage. Yet for women, as for men, employment is no guarantee of health insurance. While highly educated professional women tend to have health insurance, poorly educated women who work in the low-wage service sector often do not. Despite the growing number of women who obtain health insurance through their own employment, marriage continues to represent a significant source of insurance for a large fraction of women (Harrington Meyer and Pavalko, 1996; Moen and Roehling, 2005). As is the case with employment, marriage is no guarantee of health insurance. The smaller economic gains from marriage for some women, coupled with the current climate of marital instability, means that marriage is neither a likely nor a long-term source of insurance for many. Furthermore, historically institutionalized gender roles, occupational segregation, and wage gaps place many women at a disadvantage relative to men in terms of their own employment-based health insurance, which can increase their reliance on men as the primary economic provider in marriage (Orloff, 1993; Pascall and Lewis, 2004).
In addition to gender, minority group status constitutes a significant structural barrier to health insurance in the United States. Compared to non-Hispanic whites, African-American and Mexican-origin adults are more likely to be employed in low-wage jobs that do not offer coverage. Among the working age population, 45% of Mexican-origin adults lack coverage, compared to 24% of non-Hispanic Blacks, and 13% of non-Hispanic whites (Angel, Angel, and Lein, 2007b). Given the long history of minority exclusion from full employment in the U.S., gender-based and economic inequalities may interact with race and ethnicity to further restrict health insurance coverage for minority women, notably for women of Mexican ancestry.
In this research, we examine how employment and marital roles relate to health insurance coverage for Mexican-origin women compared with non-Hispanic white and African-American women. We focus particularly on Mexican-origin women because at all ages the Mexican-origin population has the lowest rate of health insurance coverage of any racial or ethnic group in the U.S. (Escarce and Kapur, 2006). In addition to the labor force disadvantages of the Mexican-origin population, aspects of Mexican culture associated with familism and masculine machismo may further weaken the family and labor market position of Mexican-origin women (see Vandello and Cohen, 2003). Thus, employment and marital roles may have unique implications for health insurance coverage for these women. As of yet, little empirical work has been given to how employment and marital roles combine to shape insurance coverage among women (Harrington Meyer and Pavalko, 1996). In this study then, we (a) test two competing hypotheses to determine whether employment and marital roles substitute for each other or whether they have a cumulative association with insurance coverage, and (b) evaluate whether role combinations exist for which health insurance coverage is similar for these women.
Two opposing theoretical perspectives of health insurance coverage dominate the literature. First, individualistic perspectives assert that human capital and rational choice predict insurance coverage. They claim coverage results from adequate human capital accumulation including education, job training, and job tenure, as well as individual choices to maintain an uninterrupted employment trajectory instead of reducing work hours or engaging in intermittent labor to manage family responsibilities (Becker, 1993; for a review, Seccombe, 1993). Indeed, education is often associated with private health insurance above and beyond occupational characteristics (Keene and Prokos, 2007; Seccombe, Clarke, and Coward, 1994), and intermittent labor force attachment is a barrier to stable, employment-based insurance (Moen and Roehling, 2005). However, this perspective neglects the role of structural features such as gender, race/ethnicity, and marital status in shaping labor market position and health insurance coverage (Harrington Meyer and Pavalko, 1996). Furthermore, the primary focus on individual, employment-based insurance overlooks marriage and other sources of insurance for women.
On the other hand, structural perspectives assert that an individual's position in the labor market, as opposed to human capital, determines health insurance (Seccombe, 1993; Seccombe and Beeghley, 1992). Indeed, larger companies, unions, the public sector, fulltime jobs, and higher wage occupations are more likely to offer health insurance (Fronstin, 2007). For example, 95% of those employed in the public sector are offered health insurance compared to 78% in the private sector; and only 43% of those employed between 21 and 34 hours a week are offered insurance compared to 82% of those employed fulltime (Fronstin, 2007). In addition, eligibility restrictions and costly premiums prevent many from accepting coverage when it is available. Among the 37% of workers who do not have insurance from their employer, 50% work for employers that do not offer coverage, 18% are ineligible for coverage (due to wait periods or temporary or part-time employment), and 32% choose not to be covered because they have another source of insurance or because the insurance is too expensive (Fronstin, 2007). Like individualistic perspectives, structural perspectives overlook the important roles that gender, race/ethnicity, and marriage play in shaping all sources of health insurance for women.
Perhaps most important, neither perspective sufficiently recognizes the complexity of women's lives and how their family and work roles intersect to affect their health insurance status (Harrington Meyer and Pavalko, 1996). The ways that women combine employment and family have considerable consequences for their insurance coverage (Spain and Bianchi, 1996) such that employment may be more beneficial for some women while marriage may be more beneficial for others. Unmarried women may be more compelled to work fulltime and seek jobs that offer insurance than married women who can more readily accept part-time jobs or jobs without health insurance if they have spousal coverage (Harrington Meyer and Pavalko, 1996).
Although scant attention has been given to the combination of employment and marital roles on health insurance, health research has a long history of evaluating the combination of these roles on women's physical and mental health. Traditionally, two competing hypotheses about the effect of multiple roles on health have been evaluated in this literature. The Role Strain Hypothesis asserts that engaging in multiple roles causes role overload and strain which is deleterious to women's health (Gove and Tudor, 1973). Conversely, the Role Accumulation Hypothesis asserts that the accumulation of multiple roles provides a variety of social and financial rewards that enhance health and well-being (Thoits, 1986) in an approximately additive manner. Most empirical evidence supports the latter hypothesis or a recent variant, the Role Substitution Hypothesis (Waldron, Weiss, and Hughes, 1998). This hypothesis claims that because employment and marriage provide similar benefits, they substitute for each other, and this is evidenced by their interaction in statistical models. In fact, several studies demonstrate an interaction between employment and marriage on the physical health (Arber, 1991; Waldron, Hughes, and Brooks, 1996; Waldron et al., 1998) and mental health (Ali and Avison, 1997; Krause and Markides, 1985; Thoits, 1986) of women. In general, this evidence suggests that employment is primarily beneficial for unmarried women, while marriage is primarily beneficial for unemployed women, since these roles provide similar resources for health. Furthermore, whether and how these roles interact to affect women's health may vary by race, ethnicity, and socioeconomic status (Arber, 1991; Jackson, 1997; Repetti, Matthews, and Waldron, 1989; Rushing, Ritter, and Burton, 1992). Our study builds on this literature regarding the health effects of multiple roles, and especially the research of Harrington Meyer and Pavalko (1996) to investigate competing hypotheses about the ways that employment and marital roles combine to affect health insurance coverage among women, particularly Mexican-origin women.
In recent decades, women's employment has replaced marriage as their primary source of health insurance (Short, 1998). However, race and ethnicity structure employment opportunities unequally, which results in unequal material resources such as health insurance (Heinz, 2004; Higginbotham, 1997). The employment opportunities of non-Hispanic white, African-American, and Mexican-origin women can be summarized by their employment rates, education levels, earnings, occupational characteristics, and employment continuity. Each of these factors, as discussed below, has important implications for health insurance coverage
Engaging in paid employment is a prerequisite for one's own employment-related coverage. African-American women are the most likely to be employed with 62% in the civilian labor force, followed closely by non-Hispanic white women at 59%, and finally Mexican-origin women at 53% (U.S. Census Bureau, 2006a). Less than fifty percent (48%) of foreign-born Mexican-origin women are in the labor force compared to 59% of native-born women. Multiple barriers including low education, lack of English proficiency, and lack of legal documentation constrain employment opportunities for many foreign-born women.
Education is also a strong predictor of insurance coverage in part because it enhances access to benefit-rich occupations and high incomes. The average educational attainment among Mexican-origin women is just 9.2 years among the foreign-born and 11.1 years among the native-born, much lower than the 12.9 years among African-American women and 13.7 years among non-Hispanic white women (Everett et al., 2007). Not only do foreign-born Mexican-origin women have low education levels, their returns to education are attenuated because U.S. employers tend to place less value on education obtained abroad (Duncan, Hotz, and Trejo, 2006).
Earnings are associated with health insurance in part because high-income occupations tend to offer insurance and income provides the means to pay insurance premiums. Among fulltime, year-round workers at least 15 years of age, the median incomes of African-American and Mexican-origin women were 89% and 71%, respectively, of the median income for non-Hispanic white women in 1999 (tabulation from Census 2000 Summary File 4). Differences among Mexican-origin women are again considerable as foreign-born Mexican-origin women earn roughly 73% of the income of their native-born counterparts (U.S. Census Bureau, 2002a). Low education and lack of English proficiency largely explain the lower earnings of Mexican-origin women (Duncan et al., 2006).
Contemporary occupational distributions are more favorable for non-Hispanic white and African-American women than Mexican-origin women with respect to health insurance. While non-Hispanic white women are advantaged because they tend to hold high income, professional and managerial occupations, African-American women are advantaged because they are more likely to be employed in the public sector, in large firms, and in unionized jobs. In contrast, Mexican-origin women are overrepresented in low-wage, service occupations (U.S. Census Bureau, 2006a). Because undocumented women are restricted from working in the formal sector, many find employment in the informal sector in which health insurance is not offered (Hondagneu-Sotelo, 1997).
Finally, health insurance is generally restricted to individuals who maintain continuous, fulltime employment throughout their adult years (Heinz, 2004; Moen and Roehling, 2005). This employment-based system emerged after World War II and was built on the male breadwinner model in which married men could maintain continuous labor force attachment because a non-employed spouse managed the family domain (Moen and Roehling, 2005). The system remains embedded in our employment benefit policies even though it is incongruous with contemporary employment-family life. As a consequence, women who exit and re-enter the labor force in response to family needs risk losing health insurance. African-American women have historically exhibited more continuous labor force attachment than non-Hispanic white women because they are often the sole provider for their families (Yoon, 1996). Others have similarly suggested that Mexican-origin women work out of financial necessity because the low earnings of Mexican-origin men cannot support a family (Moreno and Muller, 1996).
As women's roles have changed over the last century, marriage has become the second most common source of health insurance for women. Married women are more likely to have private insurance than unmarried women, and unmarried women are more likely to be uninsured or to rely on public insurance (Anderson and Eamon, 2004; Hahn, 1993; Harrington Meyer and Pavalko, 1996). Married women also have more continuous coverage (Short, 1998). Yet, coverage through marriage depends first on the likelihood of marriage. It is well-documented that non-Hispanic white and Mexican-origin women are more likely to be married than African-American women. In addition, marriage may be less of a guarantee of coverage for minority women. Unlike non-Hispanic white women, African-American women tend to benefit economically more from employment than marriage (Angel, Jiménez, and Angel, 2007a; Willson, 2003). Furthermore, although marriage increases the odds of having a family income above the poverty threshold for non-Hispanic white, African-American and Hispanic women, it does not raise family income to very high levels for Hispanic women (Lichter, Graefe, and Brown, 2003).
The attenuated benefits of marriage among minority women may result from differences in marriage predictors and/or marriage timing. In fact, marriage market predictors for potential spouses may operate differently for Hispanic women. For non-Hispanic white and African-American women, first marriage probabilities are positively associated with men's employment in a marriage market (Lloyd, 2006). Yet for Hispanic women, men's human capital and market characteristics do not appear to affect the probability of marriage, which may partly explain the smaller economic gain to marriage for Mexican-origin women (see Lloyd, 2006).
Earlier marriage among Mexican-origin women may also explain its attenuated economic benefit. Their early marriage pattern has been considered a paradox (Oropesa, Lichter, and Anderson, 1994) since a disadvantaged socioeconomic status is associated with delayed marriage among non-Hispanic white and Black women. Cultural explanations point to a greater pro-nuptial orientation among Mexican-origin women (Oropesa, 1996). Indeed, Hispanic middle-school girls report lower school and job aspirations and a greater desire to marry and become mothers at earlier ages than non-Hispanic white and African-American girls (East, 1998). Yet, recent structural explanations assert that impeded educational opportunities combined with high employment rates among Mexican-origin men encourage early marriage, calling into question whether this pattern represents a paradox (see Landale and Oropesa, 2007).
In addition, the presence and timing of motherhood has profound consequences for the life course trajectories of women (Spain and Bianchi, 1996). Women remain primarily responsible for childrearing, and as a consequence employed women are more likely than employed men to reduce work hours or temporarily exit the labor force in response to family demands. Not surprisingly, motherhood is associated with lower wages and fewer fringe benefits, an effect of parenthood that is opposite to that on fathers (Moen and Roehling, 2005). Partly as a result of increasingly incompatible employment and family demands, many women have responded by delaying or foregoing marriage and/or childbearing (Spain and Bianchi, 1996), although early marriage, early childbearing and larger families among Mexican-origin women remain common (Landale and Oropesa, 2007).
Yet, health insurance through marriage is possible only if one marries and stays married. Given current levels of marital instability, reliance on marriage as a source of insurance is risky. Indeed, married women insured as dependents are more likely than those insured as policyholders to lose insurance (Short, 1998). Since the 1986 Consolidated Omnibus Budget Reconciliation Act (COBRA), women can maintain their spouse's employment-based health insurance for up to 36 months after a divorce, but the costs are often prohibitive because she must pay both her portion and the employer's portion of the premium (U.S. Department of Labor, 2008). Obtaining insurance from the individual market is also an elusive alternative. Almost 90% of adults who search for individual coverage do not purchase it due to costs or because they were denied coverage (Collins et al., 2006). Unlike many financial resources, health insurance cannot be stored. Thus, current marital status, rather than marital history, is the significant predictor of coverage among women (Harrington Meyer and Pavalko, 1996).
In the following analyses we examine how employment and marital roles relate to health insurance coverage among Mexican-origin women, compared with non-Hispanic white and African-American women. Our analyses address the following research questions.
We anticipate that employment and marriage substitute for each other for non-Hispanic white women. These women and men share labor market advantages such that either marriage or employment may be sufficient for coverage. In contrast, we anticipate that employment and marriage operate additively for African-American and Mexican-origin women. African-American women may marginally increase their odds of coverage through marriage but not enough to substitute for their own employment, whereas both the employment disadvantages and smaller gains from marriage for Mexican-origin women may preclude a high likelihood of coverage from either source. Finally, we anticipate that race/ethnic disparities in coverage do not exist within every role combination since race/ethnicity may moderate the association between these roles and coverage.
We use data from the Annual Social and Economic (ASEC) Supplement to the March Current Population Survey. The ASEC provides demographic, economic, and health insurance information for a nationally representative sample of roughly 99,000 households including an oversample of the Hispanic population. Because the ASEC uses a rolling sample design, calendar months in consecutive years of the ASEC contain roughly one-half of the same households. Thus, we combine the 2004 and 2006 ASEC to ensure sufficient sample sizes of Mexican-origin and African-American women without duplicating households (U.S. Census Bureau, 2002b, 2006b). The analytic sample consists of women between 18 and 64 years of age who were not fulltime students and who reported their race and ethnicity as non-Hispanic white, non-Hispanic black (hereafter African-American), or Mexican-origin. Race and ethnicity are self-reported in the ASEC. Prior to self-reporting race, individuals were asked to report whether they were Spanish, Hispanic or Latino. Those who responded affirmatively were shown a flash card listing five nationalities (Mexican, Puerto Rican, Cuban, Central/South American, and Other Spanish) and asked to select one. We identify women as Mexican-origin if they reported a Mexican ancestry, regardless of race or nativity. We exclude widowed women due to small cell sizes among this age range.
Our dependent variable is a three-category variable indicating whether a woman had private health insurance (1), public health insurance only (2), or no health insurance of any kind (3) during the previous calendar year. Private insurance includes employment-based, privately purchased, and military health insurance such as TRICARE/CHAMPUS and CHAMPVA. Public insurance includes Medicaid, Medicare, and other state programs for low-income adults. We do not distinguish between policyholder and dependent statuses among the privately insured. Because the focus of our analysis is how different combinations of roles correspond with coverage, we use one aggregate measure that captures any source of private insurance gained from the accumulation of those roles.
Our main explanatory variables are employment status, marital status, their interaction, and family income as a potential mediator between these roles and health insurance. Like health insurance, employment status and family income refer to the previous calendar year. Marital status is ascertained at the time of survey. We analyze employment status using three dummy variables indicating fulltime, part-time, or not employed (omitted reference). We use three dummy variables to measure marital status: married, divorced, or never married (omitted reference). Family income is assessed with an income-to-poverty ratio that combines total family income and adjusts for family size and age of members. We include five dummy variables to indicate a ratio less than 1.00 (omitted reference), 1.00-1.99, 2.00-2.99, 3.00-3.99, or 4.00 and higher.
We incorporate relevant covariates consistently linked to health insurance. Age is measured with three dummy variables indicating 18-35 (omitted reference), 36-48, or 49-64 years. We include age as a categorical variable because the likelihood of insurance increases until the late 40s when, for women, it declines until Medicare eligibility (Lambrew, 2001). We capture educational attainment with four dummy variables indicating less than high school, high school degree (omitted reference), associates degree or some college, or a bachelor's degree or higher. We combine nativity and U.S. citizenship into three dummy variables identifying native-born citizens (omitted reference), naturalized citizens, or non-citizens. Health insurance rates vary between these groups since non-citizenship and foreign birth present distinct barriers to coverage. Not only do non-citizens encounter employment-related barriers, but non-citizenship is also a barrier for public insurance, while a lack of English proficiency for completing insurance forms and communicating with physicians is a common barrier for all types of health insurance, regardless of citizenship (Ku and Waidmann, 2003). We include a dummy variable indicating whether the woman has children under 18 years of age (no children under 18 is the omitted reference) since young children may correspond with public insurance among low-income women. Finally, we include self-rated health as a dummy variable indicating fair or poor health versus excellent, very good or good health (omitted reference). We control for self-rated health since adults who report fair or poor health are often unable to work outside the home, and they are more likely to rely on Medicaid (Kaiser Commission on Medicaid and the Uninsured, 2007).
To address our second research question, we create a 27-category typology of all combinations of our three race/ethnicities, three employment statuses, and three marital statuses. We analyze these categories using 27 dummy variables (omitted reference is non-Hispanic white, fulltime employed, married women). Typologies are an effective tool for analyzing role combinations when the effect of one role depends on another (Jackson, 1997; Menaghan, 1989; Rushing et al., 1992; Thoits, 1986). Our results from research question 1a will show that the association between employment and coverage depends on marital status, and vice versa, for two race/ethnic groups. Further, typologies allow us to more clearly demonstrate how race/ethnicity, employment, and marital status combine to influence coverage.
To address research questions 1a-b, we estimate nested multinomial logistic regression models stratified by race/ethnicity. Model 1 includes the main effects of employment and marital status. Model 2 adds the employment by marital status interaction. A significant, negative interaction suggests that employment and marriage are substitutes for health insurance which supports a Role Substitution Hypothesis. An insignificant interaction with significant main effects supports a Role Accumulation Hypothesis. In model 3, we control for women's age, education, self-rated health, nativity and citizenship, and whether she has minor children. In model 4, we also control for family income. Model 4 allows us to (a) incorporate information about spouses' contribution to household resources among married women, and (b) further test the Role Substitution Hypothesis (where it is supported) by examining whether income is a major substitutable resource gained through employment and marriage. To address research question 2, we estimate a single multinomial logistic regression model using the 27 dummy variables from the typology while controlling for all covariates and family income.
Table 1 displays the weighted means of the explanatory and control variables. Compared to non-Hispanic white and African-American women, Mexican-origin women have lower education levels, are less likely to be employed fulltime, and have lower family income-to-poverty ratios (yet they are as likely as African-American women to live below the poverty line). Mexican-origin women are as likely to be married as non-Hispanic white women, and more likely than African-American women. The distribution of employment and marital combinations highlights differences in normative role combinations. The modal role combination is not employed but married among Mexican-origin women (29.5%), fulltime employed and married among non-Hispanic white women (36.4%), and fulltime employed and never married among African-American women (27.1%); however the fulltime employed and married combination was one of the two most common combinations for all women. Although the table highlights the disadvantaged socioeconomic position of Mexican-origin women, they report better health than African-American women. This well-documented paradox may result from selective migration and/or positive health behaviors among less acculturated Mexican-origin women (Lopez-Gonzalez, Aravena, and Hummer, 2005), as well as deleterious health consequences of racial discrimination upon African-American women (Geronimus, 1992). Finally, the table confirms that Mexican-origin women are less likely to have any form of health insurance. Just 57.8% of Mexican-origin women report some form of health insurance, compared to 78.7% of African-American women, and 86.9% of non-Hispanic white women.
Next, we turn to descriptive analyses of public and private health insurance rates in Table 2. Mexican-origin (47.1%) women are less likely than non-Hispanic white (80.8%) and African-American (62.5%) women to have private insurance, and this gap is not compensated by public insurance. Table 2 also illustrates that employment and marriage are associated with each type of health insurance, with the strength of that association varying by race/ethnicity. For each race/ethnic group, women who are employed fulltime are more likely (tests not shown) to have private insurance and less likely to rely on public insurance, than part-time or non employed women. For each race/ethnic group, married women are more likely (tests not shown) than divorced or never married women to have private insurance with one exception: married and divorced Mexican-origin women are similarly likely to have private insurance.
We now address our first set of research questions using the models in Table 3, which are stratified by race/ethnicity. Our first question asks how are women's employment and marital roles related to their health insurance coverage, and specifically whether the combination of these roles supports a role substitution or role accumulation hypothesis. We focus here on private insurance. The main effects model(s) 1 shows that compared to non-employed women, the odds of private insurance for the fulltime employed are 3.1 times greater [exp(1.140)] among non-Hispanic white women, 4.6 times greater among African-American women, and 4.0 times greater among Mexican-origin women. Compared to never married women, the odds of private insurance for married women are 4.4, 3.8, and 2.2 times greater for non-Hispanic white, African-American, and Mexican-origin women, respectively. Employment and marriage increase the odds of private coverage for all women. However, in the main effects model, non-Hispanic white women increase their odds of private coverage more through marriage than fulltime employment, while the opposite pattern is apparent for Mexican-origin women.
Model(s) 2 in Table 3 adds the employment by marital status interaction. We find a significant, negative interaction between fulltime employment and marriage among non-Hispanic white and African-American women, which indicates these roles partially substitute for each other among these women. However, marriage increases the odds of coverage slightly more than fulltime employment for never married non-Hispanic white women. The results further suggest that these roles are close substitutes for never married African-American women since their fulltime employment (1.867) and marriage (1.802) coefficients are similar and their interaction (-.726) discounts more of either coefficient. In sharp contrast, employment and marriage operate additively for Mexican-origin women since both main effects are significant and there is no significant interaction. Fulltime employment increases the odds of coverage more than marriage. Taken together, these findings provide support for a Role Accumulation Hypothesis among Mexican-origin women and tentative support for a Role Substitution Hypothesis among non-Hispanic white and African-American women. To facilitate interpretation of model(s) 2, Figure 1 depicts the ln(odds) of private insurance estimated from these models for the different combinations of employment and marital status. It is worth noting that Mexican-origin women who are fulltime employed and divorced exhibit slightly greater than expected ln(odds) of coverage, yet the p-value for the interaction (p=.07) falls above our significance threshold.
The Role Substitution Hypothesis asserts that marriage and employment substitute for each other because they provide similar resources. We now further evaluate this hypothesis by asking whether family income is an important substitutable resource. We first control for women's covariates in model(s) 3, and then compare these coefficients to model(s) 4 that also include family income. Controlling for family income substantially reduced the fulltime employment coefficient by 45%, the marriage coefficient by 37%, and their interaction by 23% among non-Hispanic white women. The substantial reduction in these coefficients lends further support to the Role Substitution Hypothesis. For African-American women, the fulltime employment coefficient was reduced by 29%, the marriage coefficient by 23%, and their interaction by just 4%. The smaller reduction in these coefficients suggests that a non-income resource is also quite important. For Mexican-origin women, the main effect of fulltime employment was reduced by 43%, the main effect of marriage by 36%, and their (insignificant) interaction by 41%, suggesting that income is a very important mediator of these roles.
Since the combination of employment and marital roles seems to have different implications for insurance coverage for each group of women, our second research question asks whether any role combinations exist for which the odds of coverage are similar among these women. Figure 2 displays the odds ratios obtained from a single, multinomial logistic regression model that includes 27 dummy variables created from three race/ethnicities, three employment statuses, and three marital statuses (omitted reference are non-Hispanic white, fulltime employed, married women), and controls for all covariates and family income. Using the same model, we created a separate figure to display the odds of public insurance (available on request). The vertical bars in Figure 2 represent the odds ratios of private insurance for each role combination compared to the omitted reference. Stars indicate that the odds of private insurance among Mexican-origin or African-American women are significantly different from non-Hispanic white women within that role combination. Thus, Figure 2 reveals that even after controlling for demographics and family income, Mexican-origin women are less likely to have private insurance than non-Hispanic white women within most role combinations. However, three role combinations exhibit relatively similar odds of coverage. Divorced Mexican-origin women are similar to non-Hispanic white women when employed part-time, and only marginally different when employed fulltime or not employed. African-American women are as likely as non-Hispanic white women to have private insurance within all but three types: part-time employed and married, part-time employed and never married, and not employed and never married. Thus, Figure 2 demonstrates that race/ethnicity, employment, and marital status interact so that insurance disparities do not exist within every role combination.
Our study contributes to the literature on gender, race/ethnicity, and health insurance by illustrating how the combination of employment and marital roles uniquely shapes insurance coverage among Mexican-origin women, compared with non-Hispanic white and African-American women. We tested two competing hypotheses to evaluate whether these roles substitute for each other or whether they have a cumulative association with insurance coverage. Our findings indicate strong support for a Role Substitution Hypothesis for non-Hispanic white women, although marriage tends to increase the odds of private coverage somewhat more than employment. Income is a highly significant mediator through which both of these roles provide private coverage. We also find support for a Role Substitution Hypothesis for African-American women. Income is an important but less relevant mediator for these women. It is likely that non-income factors such as their greater likelihood of being employed in large firms, the public sector, and unionized jobs are more relevant mediators gained from their employment and marriage. In sharp contrast, we find support for a Role Accumulation Hypothesis for Mexican-origin women. Employment increases their odds of coverage more than marriage. Yet, neither employment, nor marriage, nor their combination assures insurance coverage for these women. Family income explained a remarkably similar and substantial proportion of the effects of employment and marriage for Mexican-origin and non-Hispanic white women, suggesting that income operates similarly as a mediator between these roles and private coverage for these women, more so than it does for African-American women.
We also observed support for our second expectation that the well-documented race/ethnic disparities in coverage do not exist within every role combination. While other multivariate analyses show a persistent health insurance disadvantage for Mexican-origin women and similar coverage between African-American and non-Hispanic white women, our findings underline that these analyses mask substantial heterogeneity within role combinations. Although disadvantaged employment characteristics contribute to low rates of private coverage among Mexican-origin women, differential gains from marriage emerged as the main contributor, even after controlling for family income. Several factors might account for this. One factor might be occupational characteristics of spouses. Mexican-origin men are more likely to work in service, construction, and agricultural occupations that have low rates of coverage. However, occupational characteristics explain a small portion of the insurance disparity between Hispanic and white males, while income differences emerge as the largest contributor by far (Monheit and Vistnes, 2000). Ancillary analyses that controlled for spouse's employment characteristics (available on request) did not reduce the significant differences in our study suggesting that, net of income, occupational differences contribute little to the disparities. A second factor might be the lower rate of acceptance of health insurance among Hispanic men. When offered insurance through employment, 76% of Hispanic males accept, versus 85% of white males and 84% of African-American males (Monheit and Vistnes, 2000). It is unclear whether their acceptance rate is lower because the costs simply cannot fit within an already strained household budget, or whether the value of health services gained through insurance are not perceived to outweigh their barriers, costs, or other reasons. Indeed, among Hispanic men and women, language barriers are an often cited reason for not seeking medical care, for receiving poor quality care including misdiagnoses and inappropriate medications, and for not completing insurance applications (see Ku and Waidmann, 2003). Finally, some foreign-born men may prefer to travel to Mexico for medical care especially since Mexico's 2003 passage of the System for Social Protection in Health reform, which has been gradually providing health insurance to all Mexican citizens with a goal of universal coverage by 2010.
Our finding that divorced women had comparable odds of coverage is noteworthy and may also indirectly help explain the disparity in coverage between married women discussed above. In previous generations, divorced women were more likely to be offered employment-based coverage than married women, perhaps because they were more compelled to seek jobs that offered coverage (Harrington Meyer and Pavalko, 1996). Additional analyses (available on request) using public sector employment, large firm size, and professional occupations as proxies for high offer rates revealed that divorced, fulltime employed Mexican-origin women were significantly more likely to work in the public sector and in large firms, and equally likely to be in professional occupations compared to their married, fulltime employed peers. In contrast, divorced non-Hispanic white and African-American women were either similar or disadvantaged to their married peers on these proxies. Thus, part of the explanation for similar coverage between divorced women may be the net result of these opposing patterns on offer rates. It is also conceivable that aspects of Mexican culture such as masculine machismo and respect play a role in acceptance rates. For example, unlike Hispanic women, Hispanic men are less likely to accept coverage compared to their white and Black peers. Thus, divorced Mexican-origin women may be similarly likely to have coverage as other divorced women because, unlike their married peers, they make their own, more favorable, decisions about accepting coverage.
Although these findings are consistent with what is known about women and the Mexican-origin population, the study has some limitations. First, our data did not contain life course factors such as marital and employment histories. Although the lack of information on marital history represents a potential shortcoming, it seems less relevant than current marital status for having health insurance (Harrington Meyer and Pavalko, 1996). Employment histories may be informative however. Second, we did not control for detailed employment or spousal characteristics other than family income. We selected family income because it is available for everyone regardless of employment or marital status, and because it is well suited for capturing the cumulative contribution of employment and marriage to test the role hypotheses. We recognize, as have others (Waldron et al., 1998), that focusing on income provides only a partial test of these hypotheses, and that employment characteristics may also help explain why these roles substitute for some women but accumulate for others. Despite these limitations, this research illustrates the utility of using role combinations as a conceptual framework for examining women's insurance coverage. In order to improve our understanding of insurance vulnerabilities among minority women we must look beyond individual characteristics to social forces that structure women's employment and family interactions.
Future research may benefit from considering whether women and their spouses were offered coverage through employment but declined it. The ASEC does not collect this data. Historically, most women accepted coverage when offered. In 1987, 86% of non-Hispanic white women, 87% of Black women, and 85% of Hispanic women accepted coverage when offered (Monheit and Vistnes, 2000). As insurance premiums have risen, acceptance rates have declined. By 1996, acceptance rates for these women dropped to 75%, 76%, and 77%, respectively. Almost three-quarters of adults who decline insurance and go without do so because of costs (Fronstin, 2007). The similarity in acceptance rates for these women, plus our control for income, should minimize the impact on our comparisons. Finally, we encourage research that builds on this role combination framework and incorporates factors such as dissatisfaction with care, language barriers, transportation barriers, and why acceptance rates between Hispanic men and women differ. As these analyses reveal, moving beyond conventional approaches and variables may be the only way to fully explain the disparities in coverage.
A dramatic retreat from marriage, accompanied by the concomitant increase in paid employment among women of all races/ethnicities, will no doubt affect their health and social welfare. The new and compelling evidence clearly points to the extreme vulnerability of stay-at-home Mexican-origin women. This marriage vulnerability has important policy implications. For minority women, health insurance disadvantage characterizes the entire life span (Angel et al., 2007b). For these women the lack of health insurance in early and mid adulthood is followed by an increased risk of lacking Medigap insurance even after they qualify for Medicare. These results, then, suggest that any federal attempts to prevent late-life hardship through marriage promotion policies may not be an effective way to guarantee Mexican-origin women's health insurance coverage. Increasingly, a woman's insurance coverage is her responsibility and social policies that are based on a marriage or male-breadwinner model must change to reflect that new reality. One way to assure coverage for women and other vulnerable groups such as the Mexican-origin population is to move beyond the current employment and family based system of health insurance toward a system like that implemented in Massachusetts in 2006. As women spend a smaller fraction of their lives married, and as benefit-rich occupations are the purview of a minority of women, stable and equitable coverage may require a system of universal and comprehensive health insurance.
This research was supported by National Institute of Child Health and Human Development grants 5 R24 HD042849 and 2 T32 HD007081 awarded to the Population Research Center at the University of Texas at Austin